The assessment of power transformers' condition is crucial for the safe and stable operation of the power grid. This paper proposes a method for assessing the condition based on the graph convolutional network and transformer state panoramic information. The aim is to address the limitations of traditional equipment condition assessment models based on deep learning. Comprehensive state assessment models often fail to consider the correlation between state quantities, while models that do consider this correlation struggle to effectively incorporate unstructured descriptive data. The method enables an intelligent and differential evaluation of the transformer's operating status by using technical performance parameters, operating performance data, and external environment data in the equipment operation and maintenance knowledge graph. Unstructured information is effectively utilized through embedded coding. The correlation between state data is considered through the application of an adjacency matrix, allowing for explicit modeling of correlations in deep learning. This approach enhances feature interaction and information supplementation among state data, resulting in objective, accurate, and interpretable state evaluation results. Experimental results demonstrate a significant improvement in accuracy, ranging from 2.38% to 7.16%, when compared to the baseline models.